Detection of Diabetic Retinopathy Images Using a Fully Convolutional Neural Network

Abstract

The objective of this paper is to develop a model for the classification of diabetic retinopathy, a prime cause for blindness that appears due to prolonged diabetes. A deep learning model based on fully convolutional neural network is developed to classify the disease from fundus image of the patient. Here, proposed neural network consists of only six convolutional layers along rectified linear unit (ReLu) activation and max pooling layer. The model trains faster as compared to traditional convolutional neural network models as the absence of fully connected layer reduces the computational complexity. The validation of the proposed model is carried out by training it with a publicly available High-Resolution Fundus (HRF) image database. The model is also compared with various existing state-of-the-art methods which shows competitive result as compared to these models. The intelligence of our model lies in its ability to re-tune weight to overcome outliers encountered in future. The proposed model works well with an accuracy of 91.66%.